The Adaptive Biquadratic Neural Network

摘要

We study the macroscopic dynamics of three-state neural network, by using information theory principles and mean-field theory. The results are expressed in terms of the macroscopic parameters. An improvement of the information properties of the network is conceived due to the bilinear and biquadratic learning rules. A self-adaptive mechanism in the synapses is analyzed. The flow diagrams display new information paths and phases.